The Column Is the Argument

There is a width at which prose stops being a picture of writing and becomes writing. You know it before you can measure it. The eye reaches the end of a line and finds the beginning of the next one without effort, without the small stumble of a return sweep that lands one line too high or too low. Below that width the text feels chopped. Above it, the reader begins to lose the thread halfway across and quietly stops reading, which they will describe later as the piece being boring.

The number most often quoted is sixty-six characters, counting spaces. It is not a law. It is the middle of a range that runs roughly from forty-five to seventy-five, and where you sit inside that range is a real design decision with real consequences for tone. Forty-eight characters reads urgent, almost breathless. Seventy-two reads considered, unhurried, slightly institutional. This page sits at sixty-eight.

What the measure is doing

A column of text is a machine for controlling the reader's eye, and it has exactly three inputs: the width of the line, the space between lines, and the size of the type. Change one and you must change the others. Wide measures need more leading, because the return sweep travels further and needs a clearer target. Narrow measures need less, or the lines drift apart into a list.

Most bad body text on the web is one of these three inputs left at a default. A 1.5 line height on a 19px serif at 34rem is not neutral — it is a choice to make the page feel a little cramped. The setting here is 1.75, which at this size gives a baseline of exactly 1.75rem, and every vertical space on the page is a multiple of it.

Typography is the craft of endowing human language with a durable visual form. Everything else in a page is an interruption of that, and interruptions must earn their place.

That last sentence is the whole brief for this layout. There is no sidebar. There is no navigation bar pinned to the top edge. There are no cards, no feature grid, no call-to-action band in a contrasting colour. If something wanted to exist on this page, it had to fit inside the measure or justify its own removal, and almost nothing justified it.

The three permitted exceptions

Purity is a posture, not a method. Three things are allowed to acknowledge the world outside the column, and each earns it by carrying information the prose cannot.

1. The spine. A one-pixel rule running the full height of the viewport, just outside the measure, that fills with accent as you scroll. It is a progress bar that refuses to be a bar. On narrow screens it moves to the screen edge rather than disappearing.
2. Hanging section numerals. Set in monospace at eleven pixels, they sit in the margin so the headings themselves stay flush with the body text. Below 46rem they drop above the heading instead of overlapping it.
3. The drop cap. One per essay, on the opening paragraph only, in the accent colour. It is the single piece of ornament, and it is doing real work: it tells you where the article begins without a label saying article begins here.

Indentation, and why the first paragraph refuses it

Paragraph breaks in print are marked by an indent. On screens they are usually marked by a gap. Using both is a duplication, and the result reads as sloppy rather than generous. This page uses indents — 1.5em, applied to every paragraph that follows another paragraph — plus a modest gap of one baseline unit, because screen reading tolerates a little more air than a printed page does.

The first paragraph after a heading is never indented. Nothing precedes it, so there is nothing to distinguish it from. This rule is four hundred years old and it is still the fastest way to make a web page look like it was set by someone rather than generated.

Measured against the alternatives

Eight sizes sounds like a lot until you count them: micro label, small, body, lead, two heading sizes, an intermediate subhead, and the drop cap. Each one has a job that no other size can do. Nothing on this page is at a size chosen because it looked about right.

On removing things

The hardest part of a layout like this is not building it. It is the week afterwards, when someone asks where the related-articles module goes, and the honest answer is that it goes on a different page. A design whose entire premise is a single measure cannot absorb a three-across grid at the bottom without becoming a design about something else.

What you get in exchange is a reading experience with an unusually high floor. There is no arrangement of viewport sizes, no zoom level, no reader-mode extension that makes this page worse, because there is nothing to rearrange. The column simply gets narrower, the numerals step above their headings, and the spine slides to the edge.

What this style is good for

• Long-form essays, criticism, and technical writing where the argument is the product.
• Documentation landing pages that want to feel authored rather than assembled.
• Changelogs, manifestos, and open letters — anything with a voice behind it.

What it is bad for

• Anything comparative. Two things side by side need two columns, and this has one.
• Marketing pages that need to survive a reader who will not read.
• Dense reference material where scanning beats reading.

Set the measure first. Derive the leading from it, the type scale from the leading, and the spacing from the type scale. Then write something worth the width. If the writing cannot carry a page with no decoration on it, no amount of decoration will fix that — it will only make the failure harder to see, which is a different thing from fixing it.

Further reading on the practice, with our notes: the case against the twelve-column default, measuring leading in real reading sessions, and why justified text still fails on the web.